The 100 members of the Senate (along with the House of Representatives) are responsible for proposing, authoring, and voting on federal legislation. Not only that, but senators also provide advice and consent on treatises as well as executive nominations. They’re even responsible for conducting oversight on all branches of the federal government. With all of these tall orders, it would make sense that their basic reading level should be good if not great, right?
In an effort to determine the reading level of each of our senators, we used Twitter’s API to scrape the most recent tweets of every senator, Republican and Democrat alike. These tweets were then fed into the Flesch-Kincaid readability tests, which automatically determined the reading-level of each tweet. We compared these numbers by individual senators, separate parties, and the overall Senate. What you’ll find may not be the set-in-stone reading level of the senators, but it’s certainly something along those lines. Keep scrolling to see what we mean.
Our study looked at the overall reading levels of all tweets between parties. Below you’ll find the parties’ tweets’ average scores, as well as the top-scoring individuals from each party. We also thought to compare the days of the week in which each party chose to chirp in the most.
Of course, there are major limitations to creating a reading level analysis as straightforward as the one above. For instance, using just one or two polysyllabic words in a short tweet could give it an unusually high grade level, while choosing to speak more casually or not using a strictly formal sentence structure in a tweet could also significantly drop a person’s score. That said, the sample size we used was large enough to create a statistically significant difference between the two parties at a 95% confidence level. This means that based on the tweets scraped, we can very confidently say that Republican senators demonstrated a higher reading level than Democratic senators.
Tweets from Republican Lisa Murkowski of Alaska – whose words we’ll see in a moment – appeared to be carrying her team. Her words earned her an average score of 11.76, significantly higher than any other senator today. She fared nearly a full half point higher than the best performing Democratic senator, Patty Murray of Washington. So what were these senators saying that was so gosh darn smart (or not)?
We have two groups of people in Washington. The first believes in more freedom. The second believes in more free stuff.— John Kennedy (@SenJohnKennedy) November 2, 2017
It is inexcusable to deny anyone necessary medical attention, and utterly immoral to force unnecessary, life-altering medical procedures, like a hysterectomy, upon them. The treatment of women described in this report is heartbreaking and monstrously inhumane.— Senator Patty Murray (@PattyMurray) September 15, 2020
In June, I introduced the bipartisan Emergency Family Stabilization Act to help organizations serving vulnerable family and youth, such as Covenant House, navigate the uncertainty of #COVID19. We must ensure these organizations are able to continue their vital services.— Sen. Lisa Murkowski (@lisamurkowski) August 28, 2020
Lisa Murkowksi, the top scorer in the Senate, recently tweeted the following:
“In June, I introduced the bipartisan Emergency Family Stabilization Act to help organizations serving vulnerable family and youth, such as Covenant House, navigate the uncertainty of #COVID19. We must ensure these organizations are able to continue their vital services.”
It’s easy to see then, why her score was significantly higher than the lowest-scoring tweet in the Senate from Brian Schatz, which simply read, “Hey. Let’s vote! #wheresMitch.”
From these anecdotes, it would appear that the scores generally were higher among those using more words. The other tweets above represent the highest and lowest scores from each political party.
Hashtags aren’t the only way to measure common topics. This next part of the study shows the most frequently used terms throughout the entire tweet and compares the results overall, as well as by party.
As far as Twitter’s site analytics are concerned, they may have a lot to thank Donald Trump for. His name seemingly “broke the internet” compared to every other topic. His name was used nearly 13,000 times in Senate tweets alone since September 23, compared to the next most common topic, COVID-19, which was mentioned just 3,941 times. This glaring distance shouts even louder when you look at the top terms, particularly among the Democratic senators. It seems that Donald Trump was the top term tweeted about by Democrat senators with one single exception – Dianne Feinstein, who tweeted more often about the coronavirus.
Democrat senators were generally more likely than Republican senators to mention the coronavirus overall, though they appear generally less fond of using the actual hashtags. They mentioned COVID-19 553 more times than Republican senators did. They also mentioned the Supreme Court and climate change significantly more often than Republican senators, while Republican interest leaned more heavily into areas like the economy (mentioned nearly 100 more times) and health care (mentioned 109 more times).
If you want to see who is really going at each other’s throats, you’ll enjoy sifting through the graph above. Each dot reveals the amount of tagging and connecting that senators have done with each other through Twitter. Considering how often “Donald Trump” is the topic of discussion among Democrats, know that these “connections” aren’t always in the most positive light. But perhaps there are some positive connections and allyships buried within. Through just a glance, it’s easy to see how heavily connected they all are digitally, perhaps without even realizing it.
Take from this study what you may – while intelligence levels aren’t necessarily best ascertained from a person’s Tweets, the sheer volume of the writing scraped here does provide a statistical difference among people and parties at a 95% confidence level, technically speaking. And tweets that were automatically rated at a higher reading level did tend to flow and read more easily.
And perhaps this study is just a reminder that even the highest educated or most politically powerful among us can still benefit from a second look at what we’ve written. Whatever you’d like to share could be more influential and better resonate with others if you just take the time to revise once or twice and really up that reading level.
This project made use of the Twitter API to extract the most recent tweets of several U.S. senators with a sample range of 4,448 (Senator Diane Feinstein) to 799 (Senator Ben Sasse). Hashtags and other basic text extractions were performed with Tweepy, a Python library for accessing the Twitter API. All reading level scores are a grade-level conversion of the Flesch-Kincaid readability tests – these were performed using the Textstat package in Python, a library that enables several basic reading level analyses. Unless noted on individual figures (e.g., figures pertaining to 2020 only), the range of dates sampled in the data is February 16, 2012 – September 23, 2020.
Readability tests such as Flesch-Kincaid are based heavily on the ratios of words to sentences and syllables to words in a given text sample. Since Twitter is largely an informal communication medium, the data presented here should be considered alongside the limitation that it can be difficult to infer a true reading level from some of the very short samples that are common on the Twitter platform. It is also important to note that these data are only estimates of the reading levels associated with the sampled tweets themselves, not the individuals or organizations that own the sampled Twitter accounts.
Numbers tell no lies, and they’re here and ready to share. You’re welcome to share the Twitter news, but you must be sure your purposes are noncommercial and you link back to this page.